645
ISSN 1364-0380 (on line) 1465-3060 (printed)
Geometry & Topology
G
T
G G TT TG T T
G
G T
T
G
T
G
T
G
T
G
T
G
GG GG T TT
Volume 8 (2004) 645{673
Published: 22 April 2004
Units of ring spectra and their traces
in algebraic K-theory
Christian Schlichtkrull
Department of Mathematics, Oslo University
PO Box 1053, Blindern
NO-0316 Oslo, Norway
Email: krull@math.uio.no
Abstract
Let GL1 (R) be the units of a commutative ring spectrum R. In this paper we
identify the composition
R : BGL1 (R) ! K(R) ! THH(R) ! Ω1 (R);
where K(R) is the algebraic K-theory and THH(R) the topological Hochschild
homology of R. As a corollary we show that classes in i−1 R not annihilated
by the stable Hopf map 2 1s (S 0 ) give rise to non-trivial classes in Ki (R) for
i 3.
AMS Classication numbers
Primary: 19D55, 55P43
Secondary: 19D10, 55P48
Keywords:
ogy
Ring spectra, algebraic K-theory, topological Hochschild homol-
Proposed: Thomas Goodwillie
Seconded: Ralph Cohen, Haynes Miller
c Geometry & Topology Publications
Received: 25 November 2003
Revised: 21 April 2004
646
1
Christian Schlichtkrull
Introduction
Given a connective (symmetric) ring spectrum R, we follow Waldhausen and
dene the units GL1 (R) to be the union of the components in Ω1 (R) that
correspond to units in the discrete ring 0 R. With this denition GL1 (R)
is a group-like monoid whose group of components equals GL1 (0 R). As in
the case of a discrete ring there is a natural map BGL1 (R) ! K(R) to the
algebraic K-theory of R. If R is a commutative discrete ring this is split by
the determinant, but the denition of the determinant does not generalize to
the setting of ring spectra and the above map is in general not split, even if
R is commutative. For example, Waldhausen shows [21] that this fails quite
badly for the sphere spectrum. However, it turns out that the notion of traces
of matrices does generalize to ring spectra. This gives rise to the trace map
tr : K(R) ! THH(R), where the target is the topological Hochschild homology
rst dened by Bökstedt [6]. The purpose of the present paper is to identify
the composition
R : BGL1 (R) ! K(R) ! THH(R) ! Ω1 (R)
tr
r
(1.1)
when R is a commutative ring spectrum. The rst two arrows are dened
for any (symmetric) ring spectrum, whereas the denition of the last map depends on R being commutative. By denition, THH(R) is the innite loop
space associated to the realization of the cyclic spectrum [k] 7! R^(k+1) with
Hochschild type structure maps. We shall use Bökstedt’s explicit denition of
the smash products R^(k+1) . If R is commutative, the degree-wise multiplication R^(k+1) ! R denes a map to the constant cyclic spectrum. This gives
rise to the innite loop map r in the denition of R .
In order to state our main result, we need the fact that GL1 (R) has the structure
of an innite loop space when R is commutative, ie, that there exists a spectrum
gl1 (R) such that Ω1 (gl1 (R)) ’ GL1 (R). (We follow the convention to use small
letters for the spectrum associated to an innite loop space written in capital
letters.) It will be convenient for our purpose to give an explicit construction
of gl1 (R) using Segal’s notion of Γ-spaces. Let 2 1s (S 0 ) denote the stable
Hopf map.
Theorem 1.2 The composite map R admits a factorization
BGL1 (R) ! GL1 (R) ! Ω1 (R);
in which the second map is the natural inclusion and the rst map is multiplication by in the sense of the following commutative diagram in the homotopy
Geometry & Topology, Volume 8 (2004)
Units of ring spectra and their traces in algebraic K-theory
647
category of spaces,
−−−−!
BGL1 (R)
?
?
y
GL1 (R)
?
?
y
Ω1 (id^)
Ω1 (gl1 (R) ^ S 1 ) −−−−−−! Ω1 (gl1 (R)):
In the case where R equals the sphere spectrum this result is due to Bökstedt
and Waldhausen [8] (with a completely dierent proof).
It is clear from the denition that there is an isomorphism of abelian groups
i gl1 (R) = i R for i 1, but since the spectrum structures are dierent this
is not an isomorphism of s (S 0 )-modules. However, using that is realized as
an unstable map : S 3 ! S 2 , it is not dicult to check that the actions of are compatible in degrees i 2. The following is then an immediate corollary
of Theorem 1.2.
Corollary 1.3 For i 3, the composition
i−1 R = i BGL1 (R) ! i K(R) ! i THH(R) ! i R
is multiplication by 2 1s (S 0 ).
It thus follows that classes in i−1 R not annihilated by give rise to non-trivial
elements in i K(R).
Example 1.4 Let R = ko, the real connective K-theory spectrum. In this case
GL1 (ko) ’ f1g BO⊗ , where ⊗ indicates that the H-space structure is the
one corresponding to tensor products of vector bundles. Using the cobration
sequence ko ! ko ! ku, [18, V.5.15], we see that
Z
= 8k (ko) ! 8k+1 (ko) = Z=2
is surjective and that
Z=2 = 8k+1 (ko) ! 8k+2 (ko) = Z=2
is an isomorphism. We conclude that for k 1,
8k+1 BBO⊗ = Z maps non-trivially to 8k+1 K(ko);
8k+2 BBO⊗ = Z=2 injects as a direct summand in 8k+2 K(ko).
Geometry & Topology, Volume 8 (2004)
648
Christian Schlichtkrull
This example is interesting in view of the attempts [1], [2], to relate algebraic Ktheory to elliptic cohomology and the chromatic ltration of homotopy theory.
Another major source for the interest in algebraic K-theory in the non-linear
setting is the relation to high dimensional manifold theory via Waldhausen’s
work on stable concordances [20].
Example 1.5 Let R = 1 (G+ ) be the suspension spectrum of a commutative
(or E1 ) group-like monoid G. By denition, the algebraic K-theory of this
s (G ), and
spectrum is Waldhausen’s A(BG). In this case, i BGL1 (R) = i−1
+
thus classes in the stable homotopy that are not annihilated by map nontrivially to i A(BG) in degrees i 3.
the algebraic K-theory of the associated
Remark 1.6 Given a discrete ring R,
Starting with a
Eilenberg-MacLane spectrum H R reduces to Quillen’s K(R).
= 0 R, the linearization map R ! H R
gives rise to a
ring spectrum R and R
bration sequence
F ! K(R) ! K(R);
where by denition F is the homotopy bre. Let SL1 (R) be the unit component
= we get a map BSL1 (R) ! F which
of GL1 (R). Using that BSL1 (R)
is important in the understanding of how algebraic K-theory behaves under
linearization.
The proof of Theorem 1.2 breaks up into two parts. The rst part is to give a
description of R in non-K-theoretical terms as the composition
BGL1 (R) ! L(BGL1 (R))
B cy GL1 (R) ! GL1 (R) Ω1 (R):
r
Here L(BGL1 (R)) denotes the free loop space of BGL1 (R) and B cy GL1 (R)
is Waldhausen’s cyclic bar construction, see Section 3. The rst map is the
inclusion of the constant loops and the map r : B cy GL1 (R) ! GL1 (R) is given
by iterated multiplication in GL1 (R). The fact that GL1 (R) is an innite loop
space ensures that it is suciently homotopy commutative for the latter map
to be well-dened.
The second part of the proof is then to show that the composite map BGL1 (R)
! GL1 (R) is multiplication by . This follows from a general analysis of how
the free loop space of an innite loop space relates to the cyclic bar construction. Let us say that a sequence of maps of based spaces F ! X ! Y is
a homotopy bration sequence if (i) the composition is constant and (ii) the
canonical map from F to the homotopy ber of the second map is a weak homotopy equivalence. (This denition is most useful if Y is connected.) Given a
Geometry & Topology, Volume 8 (2004)
Units of ring spectra and their traces in algebraic K-theory
649
well-pointed group-like topological monoid G, there is a commutative diagram
of homotopy bration sequences
G
?
?
y
−−−−! B cy G −−−−! BG
?
?
y
Ω(BG) −−−−! L(BG) −−−−! BG;
in which the lower sequence is split by the inclusion of the constant loops
BG ! L(BG). If furthermore G admits the structure of an innite loop space,
then the upper sequence has a natural splitting B cy G ! G given by the iterated
product in G. The failure of these splittings to be compatible is measured by
the fact that the composition
BG ! L(BG) ’ B cy G ! G
is multiplication by in the sense described above for GL1 (R).
The paper is as a whole fairly self-contained, and in particular we present in
Section 4 a new explicit construction of the trace map tr : K(R) ! THH(R).
This version of the trace map is used here to identify the action on BGL1 (R),
but there are many other applications of this combinatorial construction. In
Section 2 we recall the denition of symmetric ring spectra and their units
and in Section 3 we recall Waldhausen’s denition of algebraic K-theory in this
framework. The Sections 2-4 can be read as a self-contained account of the
topological trace map.
In Section 5 we explain the innite loop structure of GL1 (R) used in the formulation of Theorem 1.2, and in Section 6 we construct the splitting r : THH(R) !
Ω1 (R) and complete the rst part of the proof. Finally, in Section 7 we consider
the relationship between the free loop space and the cyclic bar construction of
an innite loop space and nish the second part of the proof.
1.1
Notation and conventions
Let T be the category of based spaces. In this paper this can be understood
as either the category of compactly generated Hausdor (or weak Hausdor)
topological spaces or the category of based simplicial sets. However, we will
usually use the topological terminology and talk about topological monoids
etc. In both cases equivalences mean weak homotopy equivalences. In the
topological case we will sometimes have to assume that base points are nondegenerate in the usual sense of being neighborhood deformation retracts.
Geometry & Topology, Volume 8 (2004)
650
Christian Schlichtkrull
We let S n denote the n-fold smash product of the circle S 1 = I=@I . By a
spectrum E we understand a sequence fEn : n 0g of based spaces together
with based maps : S 1 ^ En ! En+1 . Again this may be interpreted either in
the topological or simplicial category. A map of spectra f : E ! F is a sequence
of based maps fn : En ! Fn that commute with the structure maps. We say
that f is an equivalence if it induces an isomorphism on spectrum homotopy
groups, the latter being dened by n E = colimk n+k Ek . All spectra we
consider will be connective, ie, n E = 0 for n < 0. We shall also assume
that the spectra we consider are convergent in the sense that there exists an
unbounded, non-decreasing sequence of natural numbers fn : n 0g such that
S 1 ^ En ! En+1 is as least n + n -connected for all n. This is not a serious
restriction as any connective spectrum is equivalent to a convergent one.
2
Units of ring spectra
In this section we recall Waldhausen’s denition of the space of units associated
to a ring spectrum. We shall work in the framework of symmetric spectra and
begin by recalling the relevant denitions from [12] and, for the version with
topological spaces instead of simplicial sets, [16].
2.1
Symmetric spectra
A symmetric spectrum is a spectrum in which each of the spaces En is equipped
with a base point preserving left n -action, such that the iterated structure
maps
m : S m ^ En ! Em+n
are m n -equivariant. A symmetric ring spectrum is a symmetric spectrum
equipped with n -equivariant maps 1n : S n ! En for n 0, and m n equivariant maps m;n : Em ^ En ! Em+n for m; n 0. In order to formulate
the axioms, let n be the composite
tw
n
n;m
n : Em ^ S n −! S n ^ Em −! En+m −! Em+n ;
where tw twists the two
i n, i 7! i − n for i >
0 : S 0 ^ En ! En and
These maps are required
factors, and n;m is the (n; m)-shue i 7! i + m for
n. Notice that n is m n -equivariant. Also, let
0 : En ^ S 0 ! En be the canonical identications.
to satisfy the following relations for all l; m; n 0:
(a) 1m+n = m (S m ^ 1n ),
Geometry & Topology, Volume 8 (2004)
Units of ring spectra and their traces in algebraic K-theory
(b) m = m;n (1m ^ En );
651
n = m;n (Em ^ 1n ),
(c) l+m;n (l;m ^ En ) = l;m+n (El ^ m;n ).
Here condition (a) states that the maps 1n assemble to give a map of spectra
1 : S ! E , where S denotes the sphere spectrum. Notice that (b) and (c)
imply that
l;m+n (El ^ m ) = l+m;n (
m ^ En )
as maps El ^ S m ^ En ! El+m+n and that
l (S l ^ m;n ) = l+m;n ( l ^ En ):
These are exactly the conditions for the maps m;n to produce a map of spectra
: E ^ E ! E , where the domain is the internal smash product in the category
of symmetric spectra. Condition (b) then says that 1 is a two-sided unit, and
(c) is the condition that the multiplication is associative. (These comments on
the internal smash product are only to motivate the denitions; we shall not
make explicit use of the internal smash product in this paper.) We say that R
is commutative if the diagrams
m;n
Rm ^ Rn −−−−! Rm+n
?
?
?
?m;n
ytw
y
m;n
Rn ^ Rm −−−−! Rn+m
are commutative.
2.2
I -spaces and I -monoids
In order to dene the units of a symmetric ring spectrum we need a combinatorial framework to keep track of the suspension coordinates. Let I be the
category whose objects are the nite sets n = f1; : : : ; ng and whose morphisms
are the injective (not necessarily order preserving) maps. The empty set 0 is
an initial object. The concatenation m t n dened by letting m correspond
to the rst m elements and n to the last n elements of f1; : : : ; m + ng gives
I the structure of a symmetric monoidal category. The symmetric structure is
given by the shues m;n : m t n ! n t m.
We dene an I -space to be a functor X : I!T . Given an I -space X , we
write XhI = hocolimI X . The homotopy type of XhI can be analyzed using
the following lemma due to Bökstedt. For published versions see [15, 2.3.7]
and [9, 2.5.1]. Let Fn I be the full subcategory of I containing the objects of
cardinality at least n.
Geometry & Topology, Volume 8 (2004)
652
Christian Schlichtkrull
Lemma 2.1 (Bökstedt) Let X be an I -space and suppose that each morphism n1 ! n2 in Fn I induces a n -connected map X(n1 ) ! X(n2 ). Then,
given any object m in Fn I , the natural map X(m) ! XhI given by the
inclusion in the 0-skeleton is at least (n − 1)-connected.
Let us say that an I -space X is convergent if there exists an unbounded, nondecreasing sequence of natural numbers fn : n 0g such that any morphism
n1 ! n2 in Fn I induces a n -connected map X(n1 ) ! X(n2 ). It follows from
Bökstedt’s lemma that in this case XhI is equivalent to the usual telescope
of the sequence of spaces X(n) obtained by restricting to the natural subset
inclusions in I . In particular, XhI is the usual directed colimit of the groups
X(n) if X is convergent.
We say that an I -space X is an I -monoid if it comes equipped with an associative and unital natural transformation
m;n : X(m) X(n) ! X(m + n);
where both sides are considered functors on I 2 . The unital condition means
that the basepoint in X(0) acts as a unit and associativity means that the
identity
l;m+n (Xl m;n ) = l+m;n (l;m Xn )
holds for all l; m; n 0. By denition an I -monoid X is commutative if the
diagrams
m;n
X(m) X(n) −−−−! X(m + n)
?
?
?
?m;n
ytw
y
n;m
X(n) X(m) −−−−! X(n + m)
are commutative. If X is an I -monoid, then XhI inherits the structure of a
topological monoid. The product is given by the composition
XhI XhI = hocolim X(m) X(n) ! hocolim X(m + n) ! XhI ;
II
II
in which the last map is induced by the monoidal structure of I . We say that
X is group-like if this is the case for XhI , ie, if the monoid of components
0 XhI is a group. We will show in Section 5 that if X is commutative and
group-like, then XhI has the structure of an innite loop space.
Remark 2.2 For I -spaces X that are not convergent, the homotopy type of
XhI may well dier from that of the usual telescope. Consider for example the
I -monoid n 7! Bn . In this case the associated homotopy colimit is equivalent
Geometry & Topology, Volume 8 (2004)
653
Units of ring spectra and their traces in algebraic K-theory
to the base point component of Q(S 0 ). To see this one uses that the natural
map B1 ! hocolimI Bn induces an isomorphism on integral homology. By
the universal property of Quillen’s plus-construction and the fact that the target
is a connected H-space, it follows that the latter is equivalent to B+
1 . The
conclusion then follows from the Barratt-Priddy-Quillen-Segal Theorem. As a
second example, let R be a discrete ring and consider the I -monoid dened
by the classifying spaces BGLn (R). By an argument similar to the above,
the associated homotopy colimit is equivalent to the base point component
of the algebraic K-theory space K(R). In these examples (and many more),
evaluating the homotopy colimit over I thus has the same eect as Quillen’s
plus-construction.
2.3
Units of ring spectra
Given a symmetric ring spectrum R, the sequence of spaces Ωn (Rn ) denes an
I -space as follows. A morphism : m ! n in I induces a map Ωm (Rm ) !
Ωn (Rn ) by taking f 2 Ωm (Rm ) to the composition
−1
S l ^f
l
S n −! S n = S l ^ S m −! S l ^ Rm −! Rn −! Rn :
(2.3)
Here : n = l t m ! n is the unique permutation that is order preserving on
the rst l = n − m elements and acts as on the last m elements. The action
on S n is the usual left action. The multiplication in R gives a multiplicative
structure
m;n : Ωm (Rm ) Ωn (Rn ) ! Ωm+n (Rm+n );
f ^g
m;n
m;n (f; g) : S m ^ S n −! Rm ^ Rn −! Rm+n ;
which is commutative if R is. We let Ωn (Rn ) be the union of the components
in Ωn (Rn ) that have stable multiplicative homotopy inverses in the following
sense: For each f in Ωn (Rn ) there exists an element g 2 Ωm (Rm ) such that
n;m (f; g) and m;n (g; f ) are homotopic to the unit 1m+n in Ωm+n (Rm+n ).
We consider Ωn (Rn ) as a based space with base point 1n and restricting the
above structure maps gives an I -monoid Ωn (Rn ) . We dene
GL1 (R) = hocolim Ωn (Rn )
I
with the monoid structure explained above. If R is convergent so is the I -space
Ωn (Rn ) , and by Lemma 2.1, 0 (GL1 (R)) = GL1 (0 (R)). If furthermore R
is commutative, the general construction in section 5 will produce a spectrum
gl1 (R) such that Ω1 (gl1 (R)) ’ GL1 (R).
Geometry & Topology, Volume 8 (2004)
654
3
Christian Schlichtkrull
K-theory and cyclic K-theory of ring spectra
In this section we recall the denition of the algebraic K-theory K(R) and
the cyclic algebraic K-theory Kcy (R) of a symmetric ring spectrum R. We also
recall the inclusion of the units BGL1 (R) ! K(R). This material is due mainly
to Waldhausen. Let Mn (R) be the symmetric ring spectrum whose mth space
is Map (n+ ; n+ ^ Rm ). The multiplication resembles multiplication of n n
matrices over an ordinary ring. (In this case the \matrices" in question have at
most one non-base point entry in each column.) We let GLn (R) = GL1 (Mn (R))
with the monoid structure coming from the multiplication in Mn (R). Using
the natural maps
Map (m+ ^ S k ; m+ ^ Rk ) Map (n+ ^ S l ; n+ ^ Rl )
! Map ((m t n)+ ^ S k+l ; (m t n)+ ^ Rk+l )
we have a notion of block sum of matrices and corresponding monoid homomorphisms
GLm (R) GLn (R) ! GLm+n (R):
These homomorphisms are associative
in the obvious sense and thus the induced
‘
maps of classifying spaces give n0 BGLn (R) the structure of an associative
topological monoid. By denition K(R) is the group completion
a
K(R) = ΩB
BGLn (R) :
n0
Notice that this is the version of algebraic K-theory with 0 K(R) = Z.
The‘classifying space of the units BGL1 (R) embeds in the 1-simplices of
B ( n0 BGLn (R)) and since there is just a single 0-simplex there is an induced map
a
1
S ^ BGL1 (R)+ ! B
BGLn (R)
n0
whose adjoint is the requested map BGL1 (R) ! K(R). The image is contained
in the 1-component of K(R).
There is a variant of all this using the cyclic bar construction B cy GLn (R).
Recall that for a topological monoid G, B cy G is the realization of the cyclic
space [k] 7! Gk+1 with simplicial operators
(
(g0 ; : : : ; gi gi+1 ; : : : ; gk ); for 0 i < k
di (g0 ; : : : ; gk ) =
(gk g0 ; : : : ; gk−1 );
for i = k
si (g0 ; : : : ; gk ) = (g0 ; : : : ; gi ; 1; : : : ; gk );
Geometry & Topology, Volume 8 (2004)
for 0 i k;
Units of ring spectra and their traces in algebraic K-theory
655
and cyclic operator tk (g0 ; : : : ; gk ) = (gk ; g0 ; : : : ; gk−1 ). We refer the reader
to [13] for background material on cyclic spaces. The degree-wise projections
(g0 ; : : : ; gk ) 7! (g1 ; : : : ; gk ) dene a simplicial map p : Bcy G ! B G, and if G
is group-like and has a non-degenerate unit there results a homotopy bration
sequence
G ! B cy G ! BG:
Since Bcy G is a cyclic space its realization has a canonical action of the circle
p
group T. Consider the composite map T B cy G ! B cy G ! BG, where the
rst map is given by the T-action. Letting L(−) denote the free loop space, the
adjoint is a map B cy G ! L(BG). It is immediate from the denition that this
is a T-equivariant map when the action on L(BG) is by multiplication in T.
The following proposition is well-known and follows easily from the denition
of the T-action. We shall prove a related result in Proposition 7.1 with a proof
that can easily be adapted to the present situation.
Proposition 3.1 There is a commutative diagram
G
?
?
y
i
p
i
ev
−−−−! B cy G −−−−! BG
?
?
y
Ω(BG) −−−−! L(BG) −−−−! BG
in which the lower sequence is the usual bration sequence associated to the
evaluation at the unit element of T. If G is group-like and has a non-degenerate
unit, then the upper sequence is a homotopy bration sequence and the vertical
maps are equivalences.
Notice that the lower sequence in (3.1) is split by the inclusion of BG in L(BG)
as the constant loops. By denition the cyclic K-theory of a symmetric ring
spectrum R is given by
a
cy
cy
K (R) = ΩB
B GLn (R) :
n0
cy
The projections p :
n (R) ! BGL(R) induce a map p : K (R) ! K(R)
which has a section in the homotopy category. The quickest way to see this is
to consider the diagram of monoid homomorphisms
a
a
a
BGLn (R) −!
L(BGLn (R)) −
B cy GLn (R);
B cy GL
n0
n0
n0
where the equivalence is a consequence of Proposition 3.1. After group completion there results a well-dened homotopy class K(R) ! Kcy (R), giving a
section of p up to homotopy.
Geometry & Topology, Volume 8 (2004)
656
4
Christian Schlichtkrull
Topological Hochschild homology and the trace
map
In this section we present an explicit construction of the topological trace
map tr : K(R) ! THH(R), where the target is the topological Hochschild
homology. In order to motivate the construction we rst recall the linear
trace map for an ordinary discrete ring tr : K(R) ! HH(R) with target the
Hochschild homology of R. The latter is the realization of a cyclic abelian
group HH (R) : [k] 7! R⊗k+1 with cyclic structure maps similar to Bcy G. The
multi-trace tr : HH (Mn (R)) ! HH (R) is the cyclic map given in degree k
by
X
tr(A0 ⊗ ⊗ Ak ) =
a0sk ;s0 ⊗ ⊗ aksk−1 ;sk ; where Ai = (ais;t ):
s0 ;:::;sk
Composing with the obvious inclusions Bcy GLn (R) ! HH (Mn (R)) we get a
cyclic map
a
a
Bcy GLn (R) !
HH (Mn (R)) ! HH (R):
n0
n0
This is a monoid homomorphism with respect to block-sums of matrices on the
domain and the abelian group structure on the target. After realization and
group completion we get maps
Kcy (R) ! ΩB HH(R)
HH(R):
The linear trace map tr : K(R) ! HH(R) is the homotopy class obtained by
composing with the homotopy section K(R) ! Kcy (R).
4.1
Topological Hochschild homology
Topological Hochschild homology is obtained by replacing the tensor products
in HH (R) by smash products of spectra. We shall follow Bökstedt [6] in making
this precise. Given a sequence of symmetric spectra E 1 ; : : : ; E r , we consider
the smash product as a multi-indexed spectrum in the natural way,
(E 1 ^ ^ E r )n1 ;:::;nr = En1 1 ^ ^ Enr r :
In general an r-fold multi-indexed symmetric spectrum E = fEn1 ;:::;nr g has an
associated innite loop space
Ω1 (E) = hocolim
Ωn1 ++nr (En1 ;:::;nr ):
r
I
Geometry & Topology, Volume 8 (2004)
657
Units of ring spectra and their traces in algebraic K-theory
The functoriality underlying this denition is analogous to that in (2.3). We
shall always use the symbol Ω1 (E) in this precise way. Notice that the
monoidal structure of I r makes Ω1 a functor from multi-indexed spectra to
topological monoids. The topological Hochschild homology of a ring spectrum
R is the topological realization of the cyclic spectrum TH (R), dened in spectrum degree n by
TH (R; n) : [k] 7! Ω1 (R
^ R} ^S n ):
| ^ {z
k+1
The spectrum structure maps are dened in the obvious way involving only
the S n -factor. This construction represents Bökstedt’s solution to the problem
of how to turn the multi-indexed spectrum R^(k+1) into an equivalent singlyindexed spectrum. The cyclic structure maps are analogous to those in Bcy (G)
and HH (R). Thus for example d0 : TH1 (R) ! TH0 (R) is the composition
hocolim Ωn0 +n1 (Rn0 ^ Rn1 ^ S n ) ! hocolim Ωn0 +n1 (Rn0 +n1 ^ S n )
I2
I2
! hocolim Ωn0 (Rn0 );
I
where the rst map uses the multiplication in R and the second map is induced by the monoidal structure t : I I ! I . It follows from the version of
Bökstedt’s approximation Lemma 2.1 with I k+1 instead of I that TH(R) is
an Ω-spectrum, and we let THH(R) be the 0th space.
In order to dene the spectrum level multi-trace, we need to model the additive
structure of a spectrum in a very precise way. We next explain how this can be
done.
4.2
The cyclic Barratt-Eccles construction
Let E n be the cyclic set [k] 7! k+1
with simplicial operators
n
di (0 ; : : : ; k ) = (0 ; : : : ; i−1 ; i+1 ; : : : ; k );
0 i k;
si (0 ; : : : ; k ) = (0 ; : : : ; i−1 ; i ; i ; : : : ; k );
0 i k;
and cyclic operator tk (0 ; : : : ; k ) = (k ; 0 ; : : : ; k−1 ). We let E1 be the
cyclic Barratt-Eccles operad with nth space E n , see [3], [17, 6.5]. This is an
E1 operad in the sense that the realization En of the nth space is n -free
and contractible. We use the notation E1
for the associated functor from based
spaces to simplicial based spaces,
a
1
n
E (X) =
(4.1)
E n X = ;
n0
Geometry & Topology, Volume 8 (2004)
658
Christian Schlichtkrull
where the equivalence relation is dened as follows. Notice rst that the
correspondence n 7! E n denes a contravariant functor from I to simplicial
sets: Given a morphism : m ! n in I and 2 n , the composition has a unique factorization = () () with () : m ! n injective and
order preserving and () 2 m . In this way induces a simplicial map,
: E n ! E m ;
(0 ; : : : ; k ) 7! ( (0 ); : : : ; (k ))
and given : l ! m it is clear that () = . Secondly, given a based
space X the correspondence n 7! X n denes a covariant functor on I by
letting a morphism : m ! n act on x 2 X m by (x) = y, where
(
xi ; if (i) = j
yj =
;
if j 2
= (m):
With this notation the equivalence relation in (4.1) is generated by the relations
(e; (x)) ( (e); x)
for e 2 Ek n ; x 2 X m and : m ! n:
In other words, E1
(X) is the tensor product of the functors n 7! E n and
n
n 7! X over I , ie, the coend of the I op I -diagram Em X n , cf. [14,
IX.6]. We let E1 (X) be the realization. (Barratt and Eccles use the notation
Γ+ (X), but we want to avoid this since we also use Γ-spaces in the sense
of Segal.) We write the elements of E1
(X) as [; x] where 2 E k and
x 2 X k . Block sums of permutations give E1
(X) the structure of a simplicial
topological monoid,
[; x] [ 0 ; x0 ] = [ 0 ; (x; x0 )]:
The homotopy theoretical signicance of the functor E1 (X) is that it provides
a combinatorial model of Ω1 1 (X) for non-degenerately based connected X .
In more detail, it is proved in [3] that in the diagram
E1 (X) ! colim Ωn E1 (S n ^ X)
colim Ωn (S n ^ X);
the left hand arrow is an equivalence for connected X and the right hand arrow
is an equivalence in general.
We extend E1 to a functor on (symmetric) spectra by applying it in each
spectrum dimension, ie, E1 (E)n = E1 (En ) with structure maps
S 1 ^ E1 (En ) ! E1 (S 1 ^ En ) ! E1 (En+1 ):
Since we assume spectra to be connective and convergent, it easily follows that
the natural map E ! E1 (E) is an equivalence. Similarly, given a simplicial
spectrum, we may apply E1
degree-wise to get a bisimplicial spectrum and
then restrict to the simplicial diagonal. This is in eect what we shall do when
dening the spectrum level multi-trace.
Geometry & Topology, Volume 8 (2004)
659
Units of ring spectra and their traces in algebraic K-theory
4.3
The spectrum level multi-trace
The multi-trace for a symmetric ring spectrum R is a natural map of multiindexed spectra
tr : Mn (R) ^ ^ Mn (R) ! E1
^ R}):
k (R
| ^ {z
|
{z
}
(4.2)
k+1
k+1
Let us rst explain how to dene this when R is a spectrum of based simplicial
sets. In this case tr is based on a natural transformation
tr : Mn (X0 ) ^ ^ Mn (Xk ) ! E1
k (X0 ^ ^ Xk );
where X0 ; : : : ; Xk are based sets and Mn (Xi ) = Map (n+ ; n+ ^ Xi ). Suppose
given an element (A0 ; : : : ; Ak ) in the domain and use matrix notation to write
Ai = (xis;t ). Let D be the set of multi-indices corresponding to the non-trivial
summands in the multi-trace formula, ie,
D = f(s0 ; : : : ; sk ) : x0sk ;s0 6= ; : : : ; xksk−1 ;sk 6= g:
Since by denition the matrices have at most one non-base point entry in each
column, the projections (s0 ; : : : ; sk ) 7! si give rise to injective maps pi : D !
n. Suppose that D has cardinality m and order the elements by choosing a
bijection γ : m ! D. The composition pi γ : m ! n is injective for each i
and admits a unique factorization pi γ = i i , where i is injective and order
preserving and i 2 m . Consider the natural map
D ! X0 ^ ^ Xk ;
(s0 ; : : : ; sk ) 7! (x0sk ;s0 ; : : : ; xksk−1 ;sk )
and let x be the composition
x : m ! D ! X0 ^ ^ Xk :
The rst observation is that the element
m
[(0 ; : : : ; k ); x] 2 k+1
m m (X0 ^ ^ Xk )
is independent of the ordering γ used to dene it. By denition the multi-trace
is the image in E1
k (X0 ^ ^ Xk ),
tr(A0 ; : : : ; Ak ) = [(0 ; : : : ; k ); x] 2 E1
k (X0 ^ ^ Xk ):
(4.3)
The second observation is that because of the base point relations in the target
this construction is natural with respect to based maps in X0 ; : : : Xk .
Example 4.4 As an example to illustrate the construction we calculate
x012
x112
tr
; 1
= [(12 ; ); (x021 ; x112 ); (x012 ; x121 )];
x021 x21 where 2 2 is the non-identity element.
Geometry & Topology, Volume 8 (2004)
660
Christian Schlichtkrull
The spectrum level multi-trace (4.2) is dened by degree-wise extending the
above natural transformation to a natural transformation between functors of
simplicial sets. We then extend this to a natural transformation of multiindexed spectra by applying it in each multi-degree. This gives the required
maps
tr : Mn (Rn0 ) ^ ^ Mn (Rnk ) ! E1
k (Rn0 ^ ^ Rnk ):
In the case where R is a spectrum of topological spaces we observe that the
expression in (4.3) also makes sense if X0 ; : : : ; Xk are (non-degenerately) based
topological spaces, and we dene tr by the same formula.
4.4
The topological trace map
We dene a combinatorially enriched version TH+ (R) of topological Hochschild
homology by applying Bökstedt’s construction to the multi-indexed spectrum
on the right hand side of (4.2), ie,
1 1
TH+
^ R} ^S n )):
(R; n) : [k] 7! Ω (Ek (R
| ^ {z
k+1
This is in a natural way the cyclic diagonal of a bicyclic spectrum. The +
decoration indicates that TH+
(R; n) is a homotopy commutative cyclic monoid.
Using the natural inclusion X 7! E1
(X) we get a degree-wise equivalence
TH (R) ! TH+
(R)
and
thus
an
equivalence
of realizations TH(R) ! TH+ (R).
The spectrum level multi-trace has formal properties similar to the linear multitrace and in particular there results a cyclic map
tr : TH (Mn (R)) ! TH+
(R):
(One can show that the realization can be extended to give an equivalence of
T-equivariant spectra, but we shall not use this here.) The denition of the
topological trace map is now completely analogous to the linear case. There is
an obvious embedding of cyclic spaces Bcy GLn (R) ! THH (Mn (R)) induced
by the natural transformation
Ωn0 (Mn (Rn0 )) Ωnk (Mn (Rnk ))
! Ωn0 ++nk (Mn (Rn0 ) ^ ^ Mn (Rnk ))
(4.5)
that sends a tuple of maps to their smash product. Composing with the multitrace we get a cyclic map
a
a
Bcy GLn (R) !
THH (Mn (R)) ! THH+
(R):
n0
n0
Geometry & Topology, Volume 8 (2004)
661
Units of ring spectra and their traces in algebraic K-theory
This is a monoid homomorphism with respect to block sums of matrices on the
domain and the simplicial monoid structure on the target. After realization
and group completion we get maps
Kcy (R) ! ΩB(THH+ (R))
THH+ (R)
THH(R):
The topological trace map tr : K(R) ! THH(R) is the homotopy class obtained by composing with the homotopy section K(R) ! Kcy (R).
Remark 4.6 It is not dicult to extend this denition of the trace map to a
map of spectra or to rene it to a version of the cyclotomic trace trc : K(R) !
TC(R), cf. [7]. However, this is not the purpose of the present paper. A
construction of the trace map from a more categorical point of view has been
given by Dundas and McCarthy [11] and Dundas [10].
Letting n = 1 in (4.5) gives a map B cy GL1 (R) ! THH(R). The following
proposition is immediate from the denitions
Proposition 4.7 There is a strictly commutative diagram of spaces
B cy GL1 (R) −−−−!
?
?
y
Kcy (R)
5
THH(R)
?
?
y
(4.8)
tr
−−−−! ΩB THH+ (R):
Γ-spaces and units of commutative ring spectra
In this section we show that if R is a commutative (and convergent) ring spectrum, then GL1 (R) is the 0th space of an Ω-spectrum. The same is true for the
group-like monoid XhI associated to a commutative and group-like I -monoid
X , and we formulate the construction in this generality.
5.1
Γ-spaces
We rst recall Segal’s notion of Γ-spaces and the Anderson-Segal method for
constructing the associated homology theory. The paper by Bouseld and Friedlander [4] is the basic reference for this material. Let Γo denote the category
of nite pointed sets and pointed maps. A Γ-space is a functor A : Γo ! T
such that A() = . We say that a Γ-space is special if given pointed sets
S and T the natural map A(S _ T ) ! A(S) A(T ) is an equivalence. This
Geometry & Topology, Volume 8 (2004)
662
Christian Schlichtkrull
implies in particular that A(S 0 ) has the structure of a homotopy associative
and commutative H-space with multiplication
A(S 0 ) A(S 0 ) ’ A(S 0 _ S 0 ) ! A(S 0 ):
We say that A is very special if A(S 0 ) is group-like, ie, if the monoid of components is a group. A Γ-space A extends to a functor on the category of pointed
simplicial sets in a two stage procedure. First A is extended to the category
of all pointed sets by forcing it to commute with colimits. Given a simplicial
set X we then apply A degree-wise to get a simplicial space [k] 7! A(Xk ) with
realization A(X). The main result is that if A is very special then the resulting
functor is a homology theory: Applying A to a cobration sequence of pointed
simplicial sets X ! Y ! Y =X gives a homotopy bration sequence
A(X) ! A(Y ) ! A(Y =X)
in the sense that the inclusion of A(X) in the homotopy ber of the second map
is an equivalence. In particular, a very special Γ-space gives rise to a symmetric
Ω-spectrum fA(S n ) : n 0g, in which the structure maps are the realizations
of the obvious (multi)-simplicial maps S1 ^ A(Sn ) ! A(Sn+1 ).
5.2
Γ-spaces associated to commutative I -monoids
In order to motivate the construction we recall the denition of the Γ-space
associated to a commutative topological monoid G. Given a nite based set S ,
let S be the subset obtained by excluding the base point. Then G(S) = GS ,
and a based map : S ! T induces a map G(S) ! G(T ) by multiplying the
elements in G indexed by −1 ftg for each t 2 T .
Implementing this idea for a commutative I -monoid requires some preparation.
be the category of subsets and inclusions in S .
Given S as above, let P(S)
by letting
A based map : S ! T induces a functor : P(T) ! P(S)
−1
(U ) = (U ) for U T . We dene a category D(S) of S -indexed sum
! I that takes
diagrams in I as follows. An object is a functor : P(S)
disjoint unions to coproducts of nite sets, ie, if U \ V = ;, then the diagram
U ! U [V
V represents U [V as a coproduct of nite sets. (The category
I itself does of course not have coproducts.) Notice in particular that ; = 0.
A morphisms in D(S) is a natural transformations of functors (not necessarily
an isomorphism). This construction is clearly functorial in Γo : A based map
: S ! T induces a functor : D(S) ! D(T ) by letting = . Notice
that an object in D(S) is determined by
its values s for s 2 S ;
Geometry & Topology, Volume 8 (2004)
663
Units of ring spectra and their traces in algebraic K-theory
a choice of injective map s ! U whenever s 2 U such that the induced
map ts2U s ! U (with any ordering of the summands) is a bijection.
Restricting to the one-point subsets of S gives a functor S : D(S) ! I S ,
where the latter denotes the product category indexed by S (we let I ; denote
the one-point category). This is an equivalence of categories, and specifying an
ordering of S gives a canonical choice of an inverse functor I S ! D(S), using
S
the monoidal structure of I . Notice however, that I is not functorial in Γo
as is the case for D(S).
Lemma 5.1 Given a functor Y : I S ! T , the natural map
hocolim Y S ! hocolim
Y
D(S)
IS
induced by S is an equivalence.
Proof By the conality criterion in [5, XI.9.2] (or rather its dual version) it
suces to check that for any object a 2 I S , the category (a # S ) of objects
under a is contractible. But this is clear since this category has an initial
object.
Let now X be a commutative I -monoid and consider the I S -diagram X S
dened by
Y
7!
fns : s 2 Sg
X(ns ):
s2S
For S = ; this should be interpreted as the one-point space. We use the
notation X(S) for the D(S)-diagram obtained by composing with S . With
this denition X(S) is functorial in S is the sense that a based map : S ! T
gives rise to a natural transformation of D(S) diagrams
X(S) ! X(T ) :
In order to see this x an object in D(S) and choose an ordering of the
subsets Ut = −1 (t) for each t 2 T . The map in question is then a product
over T of maps of the form
Y
G
X(s ) ! X(
s ) ! X(Ut );
s2Ut
s2Ut
where the rst arrow comes from the multiplication in X and the second arrow
is induced by the bijection ts2Ut s ! Ut determined by the sum diagram .
The main point is that since X is commutative the composite map does not
depend on the ordering of Ut used to dene it.
Geometry & Topology, Volume 8 (2004)
664
Christian Schlichtkrull
By denition the Γ-space associated to X is given by
XhI (S) = hocolim X(S):
D(S)
Given a based map : S ! T , the induced map : XhI (S) ! XhI (T ) is the
composition
hocolim X(S) ! hocolim X(T ) ! hocolim X(T );
D(S)
D(T )
D(T )
where the rst map is induced by the above natural transformation and the
second map is the map of homotopy colimits determined by the functor .
It follows immediately from the denition that XhI (S 0 ) = XhI . In order to
compare XhI (S 1 ) to the usual bar construction of XhI we specify an ordering
of the k -simplices in S1 by letting
uj = (0; : : : ; 0; 1; : : : ; 1);
| {z }
for j = 0; : : : ; k:
(5.2)
j
Then u0 is the base point and Sk1 = fu1 ; : : : uk g.
Proposition 5.3 The Γ-space associated to a commutative I -monoid X is
always special and is very special if and only if the underlying monoid XhI is
group-like. In general there is a natural equivalence BXhI ! XhI (S 1 ).
Proof Using Lemma 5.1 we get an equivalence
XhI (S) = hocolim X(S) ! hocolim
XS =
D(S)
IS
Y
XhI ;
s2S
which is the condition for XhI to be special. The statement about being very
special follows from the denition. In order to dene the equivalence we use the
ordering of the simplices of S1 given by (5.2). As noted earlier this ordering
determines an equivalence
hocolim X(n1 ) X(nk ) ! hocolim X(Sk1 )
Ik
D(Sk1 )
using the monoidal structure of I . Identifying the left hand side with the k simplices of B XhI we get a simplicial map B XhI ! XhI (S1 ). Since this is
an equivalence in each simplicial degree its realization is also an equivalence as
required.
Geometry & Topology, Volume 8 (2004)
665
Units of ring spectra and their traces in algebraic K-theory
Remark 5.4 This construction of Γ-spaces based on the category D(S) diers
from that of Segal [19, x2] in that we allow all natural transformations, not
only the natural isomorphisms. Consequently, our denition of the Γ-space
associated to a commutative I -monoid takes into account all the maps X(m) !
X(n) induced by morphisms in I . As an example, consider the commutative
I -monoid given by the classifying spaces BO(n) of the orthogonal groups.
In this‘case Segal’s construction produces a special Γ-space with underlying
space BO(n), whereas our construction produces a very special Γ-space with
underlying space hocolimI BO(n) ’ BO. The last equivalence follows from
Bökstedt’s Lemma 2.1. Thus the two constructions respectively produce models
of the (−1)-connected and 0-connected topological K-theory spectrum.
Denition 5.5 Given a commutative (and convergent) symmetric ring spectrum R, let GL1 (R) be the Γ-space associated to the I -monoid Ωn (Rn ) considered in Section 2.3, and let gl1 (R) be the associated spectrum.
It will always be clear from the context whether GL1 (R) denotes a Γ-space as
above or the underlying group-like monoid as in Section 2.
6
Commutative ring spectra and splittings
Let R be a commutative symmetric ring spectrum. In this section we show
that the natural inclusions GL1 (R) ! B cy GL1 (R) and Ω1 (R) ! THH(R)
have compatible left inverses in the homotopy category, where by compatible
we mean that these splittings are related by a homotopy commutative diagram
r
B cy GL1 (R) −−−−! GL1 (R)
?
?
?
?
y
y
(6.1)
r
We then dene R
THH(R) −−−−! Ω1 (R):
to be the composite homotopy class
R : BGL1 (R) ! K(R) ! THH(R) ! Ω1 (R):
tr
r
Using the diagrams (4.8) and (6.1) we get an alternative description as follows.
Proposition 6.2 The homotopy class R is represented by the composition
BGL1 (R) ! L(BGL1 (R))
r
B cy GL1 (R) ! GL1 (R);
(where the rst map is the inclusion of the constant loops), followed by the
inclusion of GL1 (R) in Ω1 (R).
Geometry & Topology, Volume 8 (2004)
666
Christian Schlichtkrull
This concludes the rst part of the proof of Theorem 1.2. In order to motivate
the construction, consider the cyclic bar construction of a commutative monoid
G. In this case the inclusion G ! B cy (G) is split by degree-wise multiplication
in G. This can also be expressed in terms of the Γ-space associated to G: The
sequence G ! B cy G ! BG is the eect of evaluating G on the cobration
1
sequence S 0 ! S+
! S1 , and the splitting is induced by the projection
1
0
S+ ! S that maps S1 to the non-base point in S 0 . Let now GL1 (R) be the
Γ-space dened in Denition 5.5. The next Lemma shows that we may replace
1 ) up to homotopy.
B cy GL1 (R) by GL1 (R)(S+
Proposition 6.3 Let X be a convergent and commutative I -monoid. Then
there exists a space W cy and equivalences
B cy XhI ! W cy
1
XhI (S+
):
Proof Let Wcy be the cyclic space
[k] 7!
hocolim X(n0 t 0 ) X(nk t k );
1 )
I k+1 D(Sk+
1 ) and we write = . We have functors
where is an object of D(Sk+
i
ui
k+1
k+1
1
1
I
!I
D(Sk+ )
D(Sk+ ) obtained by xing the initial object in one
of the factors. The cyclic structure of Wcy is the obvious one such that the
induced maps
1
Bcy XhI ! Wcy
XhI (S+
)
become maps of cyclic spaces. Since we assume that X is convergent, it follows
from Bökstedt’s approximation lemma 2.1 (with I k+1 instead of I ), that these
maps are equivalences in each simplicial degree. After realization we thus get
1 ).
a pair of equivalences relating B cy XhI and XhI (S+
Remark 6.4 The condition that X be convergent is necessary for the argument in Proposition 6.3, since otherwise the map
hocolim X(m) ! hocolim X(m t n)
I
II
induced by the functor I ! I 2 , m 7! (m; 0) need not be an equivalence. The
I -monoid X(n) = X n considered in Section 4 provides a counter example. It
should also be noted that the construction of the simplicial map B XhI !
XhI (S1 ) in the proof of Proposition 5.3 cannot be applied to give a cyclic map
1 ).
Bcy XhI ! XhI (S+
Geometry & Topology, Volume 8 (2004)
667
Units of ring spectra and their traces in algebraic K-theory
Using the above equivalences, we dene the splitting r to be the composite
homotopy class
1
r : B cy GL1 (R) ’ GL1 (R)(S+
) ! GL1 (R);
1 ! S0 .
where the last map is induced by the projection S+
We next consider a version TH0 (R) of topological Hochschild homology that
1 ) relates to B cy GL (R). By denition this is
relates to TH(R) as GL1 (R)(S+
1
the realization of the cyclic spectrum
TH0 (R; n) : [k] 7! hocolim Ω0 ttk (R0 ^ ^ Rk ^ S n );
1 )
D(Sk+
1 ) and we again write = . The cyclic
where denotes an object in D(Sk+
i
ui
1 ). For example, in spectrum
structure maps are dened as for GL1 (R)(S+
degree zero, d0 : THH01 (R) ! THH00 (R) is the composition
hocolim Ω0 t1 (R0 ^ R1 ) ! hocolim Ω01 (R01 ) ! hocolim Ω 0 (R 0 ):
1 )
2D(S1+
1 )
2D(S0+
1 )
2D(S1+
1 ) and the rst map is induced
Here 0 ! 01
1 denotes an object in D(S1+
by the natural transformation that takes f 2 Ω0 t1 (R0 ^ R1 ) to the element
in Ω01 (R01 ) given by the composition
−1
f
0 1
S 01 −! S 0 t1 = S 0 ^ S 1 −! R0 ^ R1 −!
R0 t1 −! R01 ;
;
where : 0 t 1 ! 01 is the bijection determined by . The second map
1 ) ! D(S 1 ). With this
is induced by the natural transformation d0 : D(S1+
0+
denition we have the equality
d0 = d1 : TH01 (R) ! TH00 (R)
and consequently the iterated boundary maps give a well-dened cyclic map
r : TH0 (R) ! TH00 (R), where the target is considered a constant cyclic spectrum. In spectrum degree zero we thus get a cyclic map of spaces
r : THH0 (R) ! Ω1 (R):
The next proposition is the analogue of Proposition 6.3.
Proposition 6.5 The spectra TH(R) and TH0 (R) are related by a pair of
equivalences.
Proof Letting Wcy = fWcy (n) : n 0g denote the cyclic spectrum
[k] 7!
hocolim Ωn0 t0 ttnk tk (Rn0 t0 ^ ^ Rnk tk ^ S n );
1 )
I k+1 D(Sk+
the proof proceeds exactly like the proof of Proposition 6.3.
Geometry & Topology, Volume 8 (2004)
668
Christian Schlichtkrull
As in the denition of the trace map we consider the transformation
Ω0 (R0 ) Ω(Rk ) ! Ω0 ttk (R0 ^ ^ Rk )
that sends a tuple of maps to their smash product. Viewing these maps as
1 )-diagrams we get a cyclic map
natural transformations of D(Sk+
1
GL1 (R)(S+
) ! THH0 (R):
It follows immediately from the denitions that this map is compatible with the
splittings of B cy GL1 (R) and THH0 (R) in the sense of the following proposition.
Proposition 6.6 There is a strictly commutative diagram of spaces
1) −
GL1 (R)(S+
−−−! GL1 (R)
?
?
?
?
y
y
THH0 (R)
−−−−! Ω1 (R):
The homotopy commutative diagram (6.1) in the beginning of this section is
derived from this using the equivalences in Proposition 6.3 and Proposition 6.5.
7
The Hopf map and free loops on innite loop spaces
In this section we nish the proof of Theorem 1.2 by showing that the composite
homotopy class
BGL1 (R) ! L(BGL1 (R))
r
B cy GL1 (R) ! GL1 (R)
is multiplication by in the sense explained in the introduction. More generally,
let G be a very special Γ-space and let g = fG(S n ) : n 0g be the associated
Ω-spectrum. By [4, 4.1], the Γ-space G is determined by g in the sense that
the diagram
G(X) ! hocolim Ωn G(S n ^ X)
hocolim Ωn (G(S n ) ^ X)) = Ω1 (g ^ X)
species a natural equivalence G(X) ’ Ω1 (g ^ X). Evaluating G on the based
1 gives a cyclic space G(S 1 ). The realization G(S 1 ) then has
cyclic set S+
+
+
a T-action and, as in the case of the cyclic bar construction, we consider the
composite map
1
1
T G(S+
) ! G(S+
) ! G(S 1 )
1 ) ! L(G(S 1 )). In the next proposition we analyze the homowith adjoint G(S+
topy bration sequence obtained by evaluating G on the cobration sequence
1 ! S1 .
S 0 ! S+
Geometry & Topology, Volume 8 (2004)
Units of ring spectra and their traces in algebraic K-theory
669
Proposition 7.1 There is a commutative diagram of homotopy bration sequences
1)
G(S 0 ) −−−−! G(S+
−−−−! G(S 1 )
?
?
?
?
(7.2)
y
y
Ω(G(S 1 )) −−−−! L(G(S 1 )) −−−−! G(S 1 )
in which the the vertical maps are equivalences.
Proof The commutativity of the right hand square is immediate since we
evaluate a loop at the unit element of T. In order to prove commutativity of
the left hand square we recall that for any cyclic space X , the T-action on the
zero simplices X0 jX j has the following description. If x is an element of
X0 and u 2 T,
u x = [t1 s0 x; u] 2 jX j:
Here t1 is the cyclic operator in degree one and we make the identication
T = 1 =@1 . Using this, it is easy to check that the composition
1
1
T G(S 0 ) ! T G(S+
) ! G(S+
) ! G(S 1 )
is given by (u; x) 7! [x0 ; u], where x0 2 G(S11 ) is the image of x under the homeomorphism induced by the based bijection S 0 ! S11 . The above composition
clearly equals the composition
T G(S 0 ) ! S 1 ^ G(S 0 ) ! G(S 1 );
which shows that the left hand square in the diagram is also commutative.
1 )) ! G(S 1 ) in the lower sequence is split by
In Diagram (7.2) the map L(G(S+
1 ) in the upper
the inclusion of the constant loops, and the map G(S 0 ) ! G(S+
1
sequence is split by evaluating G on the projection r : S+ ! S 0 that maps
S1 to the non-base point of S 0 . The next proposition expresses the fact that
these splittings are not compatible in general. As usual 2 1s (S 0 ) denotes the
stable Hopf map.
Proposition 7.3 Using the natural equivalences Ω1 (g ^ S 1 ) ’ G(S 1 ) and
Ω1 (g) ’ G(S 0 ), the composite homotopy class
G(S 1 ) ! L(G(S 1 ))
is given by Ω1 (g ^ ).
Geometry & Topology, Volume 8 (2004)
r
1
G(S+
) ! G(S 0 );
670
Christian Schlichtkrull
Proof We rst observe that (7.2) is in fact a diagram of innite loop spaces
and innite loop maps, and that as such it is equivalent to the following diagram
of spectra
g ^ S0
?
?
’y
−−−−!
1
g ^ S+
?
?
’y −−−−!
g ^ S1
1 ; g ^ S 1) −
F (S 1 ; g ^ S 1 ) −−−−! F (S+
−−−! F (S 0 ; g ^ S 1 ):
Here F (−; g ^ S 1 ) is the obvious function spectrum and the upper and lower
1 ! S 1 . These cobration
cobration sequences are both induced from S 0 ! S+
1 !
sequences have canonical stable splittings induced by the projection r : S+
1 . The vertical map in the
S 0 and the associated stable section s : S 1 ! S+
middle is the adjoint of
1
1
1
S+
^ g ^ S+
! g ^ S+
! g ^ S 1;
where the rst map uses the action of S 1 on itself given by the group structure.
It is clear that the above diagram is equivalent to the one obtained by smashing
g with
1
S0
−−−−!
S+
−−−−!
S1
?
?
?
?
’y
’y 1 ; S1) −
F (S 1 ; S 1 ) −−−−! F (S+
−−−! F (S 0 ; S 1 ):
Here the denition of is analogous to the denition given above. We must
prove that the stable map
r
r
1
1
S 1 = F (S 0 ; S 1 ) ! F (S+
; S 1 ) ’ S+
! S0
represents . Using the canonical splittings to represent −1 as a 2 2 matrix,
this composition represents the o diagonal term. It is therefore the negative
of the composite stable map starting in the upper right corner of the diagram,
s
s
1
1
S 1 ! S+
! F (S+
; S 1 ) ! F (S 1 ; S 1 ):
The adjoint of this is the stable map
s^s
1
1
1
S 1 ^ S 1 ! S+
^ S+
! S+
! S1;
where the second map is the group multiplication in S 1 . It is well-known that
this composition represents . For example, one can see this by considering the
1 , whose domain is the unreduced suspension
e 1 ! S 1 ^S+
equivariant splitting S
1
of S , and then use that the map of homotopy colimits induced by the diagram
S1 S1 ! ! S1 ! Geometry & Topology, Volume 8 (2004)
671
Units of ring spectra and their traces in algebraic K-theory
represents a generator of 3 (S 2 ), cf. [22, XI.4]. The result now follows since has order two.
Proof of Theorem 1.2 The only thing left to prove is that the homotopy
class BGL1 (R) ! GL1 (R) considered in Proposition 6.2 agrees with the one
in Proposition 7.3 when the Γ-space in question is GL1 (R) and we use the
canonical equivalence BGL1 (R) ! GL1 (R)(S 1 ), cf. Proposition 5.3. Let W
be the realization of the simplicial space
[k] 7! hocolim X(n1 t 1 ) X(nk t k ):
I k D(Sk1 )
Since the space W cy considered in Proposition 6.3 is the realization of a cyclic
space it has a T-action, and the adjoint of the composition
T W cy ! W cy ! W
gives an equivalence W cy ! L(W ). It easily follows that we have a commutative diagram of equivalences
−−−−
−−−− L(GL1 (R)(S 1 )):
B cy GL1 (R) −−−−! W cy
?
?
?
?
y
y
L(BGL1 (R)) −−−−! L(W )
1)
GL1 (R)(S+
?
?
y
It thus suces to check that the homotopy class dened by the diagram
BGL1 (R) ! W
GL1 (R)(S 1 )
is compatible with the canonical equivalence BGL1 (R) ! GL1 (R)(S 1 ). We do
this by exhibiting an explicit homotopy inverse of the equivalence GL1 (R)(S 1 )
! W . Let W 0 be the realization of the simplicial space
[k] 7! hocolim X(
D(Sk1 )D(Sk1 )
1
t 1 ) X(
k
t k ):
Ordering the k -simplices Sk1 as in the proof of 5.3 gives an equivalence W !
W 0 . Moreover, using the monoidal structure of I we get a functor D(Sk1 ) D(Sk1 ) ! D(Sk1 ). Varying k this is a transformation of simplicial categories and
consequently there is an induced simplicial map W0 ! GL1 (R)(S1 ). It is easy
to check that the composition W ! W 0 ! GL1 (R)(S 1 ) is a homotopy inverse
of the map in question and that the composition with BGL1 (R) ! W is the
canonical equivalence. This completes the proof.
Remark 7.4 The discussion in the above proof can be generalized to any
1 ) are
commutative I -monoid X . If X is convergent, B cy XhI and XhI (S+
Geometry & Topology, Volume 8 (2004)
672
Christian Schlichtkrull
related by cyclic equivalences as in Proposition 6.3, and if X is group-like we
get equivalences relating the same spaces by comparing them to the relevant free
loop spaces. In case X is both convergent and group-like, the two equivalences
agree by an argument similar to the above.
Acknowledgement The author was partially supported by a grant from the
NSF.
References
[1] C Ausoni, J Rognes, Algebraic K-theory of topological K-theory, Acta. Math.
188 (2002) 1{39
[2] N A Baas, B I Dundas, J Rognes, Two-vector bundles and forms of elliptic
cohomology, to appear in: \Topology, Geometry and Quantum Field Theory
(Proceedings of the Segal Birthday Conference)", Cambridge University Press
[3] M Barratt, P Eccles, Γ+ -structures-I:A free group functor for stable homotopy theory, Topology 13 (1974) 25{45
[4] A K Bouseld, E M Friedlander, Homotopy theory of Γ-spaces, spectra,
and bisimplicial sets, from: \Geometric applications of homotopy theory (Proc.
Conf. Evanstone, Ill, 1977) II", (M G Barratt, M E Mahowald editors) Springer
Lecture Notes in Math. vol. 658, Springer, Berlin (1978) 80-130
[5] A K Bouseld, D M Kan, Homotopy limits, completions and localizations,
Springer Lecture Notes in Math. Vol. 304, Springer, Berlin (1972)
[6] M Bökstedt, Topological Hochschild homology, preprint, Bielefeld (1985)
[7] M Bökstedt, W C Hsiang, I Madsen, The cyclotomic trace and algebraic
K-theory of spaces, Invent. Math. 111 (1993) 465{540
[8] M Bökstedt, F Waldhausen, The map BSG ! A() ! Q(S 0 ), from: \Algebraic topology and algebraic K-Theory, (Princeton, N. J, 1983)" Ann. of Math.
Stud. 113, Princeton University Press (1987) 418{431
[9] M Brun, Topological Hochschild homology of Z=pn , J. Pure Appl. Algebra 148
(2000) 29{76
[10] B Dundas, The cyclotomic trace for symmetric monoidal categories, from:
\Geometry and topology: Aarhus (1996)" Contemp. Math. 258 (2000) 121{143
[11] B Dundas, R McCarthy, Topological Hochschild homology of ring functors
and exact categories, J. Pure Appl. Algebra 109 (1996) 231{294
[12] M Hovey, B Shipley, J Smith, Symmetric spectra, J. Amer. Math. Soc. 13
(2000) 149{208
[13] J D S Jones, Cyclic homology and equivariant homology, Invent. Math. 87
(1987) 403{423
Geometry & Topology, Volume 8 (2004)
Units of ring spectra and their traces in algebraic K-theory
673
[14] S Mac Lane, Categories for the working mathematician, Graduate Texts in
Mathematics, vol. 5, Springer-Verlag (1971)
[15] I Madsen, Algebraic K-theory and traces, from: \Current developments in
mathematics", Internat. Press, Cambridge, MA (1994) 191{321
[16] M A Mandell, J P May, S Schwede, B Shipley, Model categories of diagram
spectra, Proc. London Math. Soc. 82 (2001) 441{512
[17] J P May, The geometry of iterated loop spaces, Springer Lecture Notes in Math.
vol. 271, Springer, Berlin (1972)
[18] J P May, F Quinn, N Ray, with contributions by J Tornehave, E1 ring
spaces and E1 ring spectra, Lecture Notes in Math. vol. 577, Springer, Berlin
(1977)
[19] G Segal, Categories and cohomology theories, Topology 13 (1974) 293-312
[20] F Waldhausen, Algebraic K-theory of topological spaces I, from: \Algebraic
and geometric topology (Proc. Symp. Pure Math. Stanford Calif. 1976) Part 1",
Proc. Sympos. Pure Math. vol. 32 (1978) 35{60
[21] F Waldhausen, Algebraic K-theory of spaces, a manifold approach, from:
\Current trends in algebraic topology, Part 1 (London, Ont, 1961)" CMS Conf
Proc. 2 (1982) 141{184
[22] G W Whitehead, Elements of homotopy theory, Graduate Text in Mathematics, vol. 61, Springer-Verlag (1978)
Geometry & Topology, Volume 8 (2004)